Description

In JetBrains Datalore before 2026.3 command execution was possible in the database task container via JVM options injection

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "6.3",
  "pubDate": "2026-11-06T11:17:06.907Z",
  "pubdate": "2026-11-06T11:17:06.907Z",
  "executiveSummary": "A critical command execution vulnerability exists within JetBrains Datalore versions prior to 2026.3, stemming from improper validation of user-supplied configuration parameters.\nThe vulnerability allows an attacker to perform JVM options injection during the initialization of the database task container.\nBy manipulating these options, an attacker can achieve arbitrary command execution within the context of the containerized environment.\nThis flaw represents a significant security risk, as it permits unauthorized remote code execution, potentially leading to full compromise of the affected database task container.\nThe impact includes the ability to execute malicious processes, exfiltrate sensitive data, or move laterally within the containerized infrastructure.\nSuccessful exploitation requires the ability to influence the configuration settings that define the JVM startup parameters for the database task, effectively bypassing existing security boundaries established for container isolation.",
  "technicalDetails": "The vulnerability is rooted in a flaw regarding the handling and sanitization of user-provided configuration inputs that are subsequently passed to the Java Virtual Machine (JVM) startup command for database task containers.\nIn affected versions of JetBrains Datalore, the application fails to adequately validate or restrict parameters provided to the JVM, enabling an attacker to inject arbitrary JVM command-line flags.\nThe attack flow involves an adversary providing a specially crafted payload—designed to leverage specific JVM options—through configuration interfaces that influence the startup environment of the container.\nBy injecting malicious options (such as those that leverage agents, system property overrides, or diagnostic tools), an attacker can redirect the execution flow of the JVM process to execute arbitrary system-level commands.\nBecause the database task container is designed to facilitate specific operations, the improper handling of these flags allows for the escape of expected functional constraints, granting the attacker the ability to execute commands with the privileges assigned to the containerized service.\nThe vulnerable component resides within the orchestration logic responsible for spawning and configuring database task containers. The lack of an allowlist or strict pattern matching for JVM startup arguments permits the inclusion of unauthorized flags.\nExploitation does not necessarily require deep knowledge of the underlying infrastructure, provided the attacker has access to configure the database task settings. Once the injected JVM options are processed during the container startup sequence, the malicious payload is executed in the runtime environment of the target process.\nPost-exploitation, the impact is severe; the attacker gains persistent command execution capabilities within the container. This level of access facilitates further unauthorized actions, including the potential to access database credentials stored in memory, environment variables containing sensitive configuration data, or internal networking components available to the container, thereby undermining the confidentiality, integrity, and availability of the Datalore deployment."
}